Move embeddings for discover section in assets

mobile/apps/photos/lib/models/ml/discover/prompt.dart

@@ -1,4 +1,3 @@
-// https://discover.ente.io/v1.json
 class Prompt {
   final String id;
   final int position;
@@ -9,6 +8,17 @@ class Prompt {
   final bool showVideo;
   final bool recentFirst;
 
+  Prompt({
+    String? id,
+    this.position = 0,
+    required this.query,
+    required this.minScore,
+    required this.minSize,
+    required this.title,
+    this.showVideo = true,
+    this.recentFirst = false,
+  }) : id = id ?? title;
+
   // fromJson
   Prompt.fromJson(Map<String, dynamic> json)
       : id = json['id'] ?? json['title'],

mobile/apps/photos/lib/services/machine_learning/semantic_search/semantic_search_service.dart

@@ -193,22 +193,13 @@ class SemanticSearchService {
     return results;
   }
 
-  Future<Map<String, List<int>>> getMatchingFileIDs(
+  Future<Map<String, List<int>>> getMatchingFileIDsWithEmbeddings(
+    Map<String, List<double>> queryToEmbedding,
     Map<String, double> queryToScore,
   ) async {
-    final textEmbeddings = <String, List<double>>{};
-    final minimumSimilarityMap = <String, double>{};
-    for (final entry in queryToScore.entries) {
-      final query = entry.key;
-      final score = entry.value;
-      final textEmbedding = await _getTextEmbedding(query);
-      textEmbeddings[query] = textEmbedding;
-      minimumSimilarityMap[query] = score;
-    }
-
     final queryResults = await _getSimilarities(
-      textEmbeddings,
-      minimumSimilarityMap: minimumSimilarityMap,
+      queryToEmbedding,
+      minimumSimilarityMap: queryToScore,
     );
     final result = <String, List<int>>{};
     for (final entry in queryResults.entries) {

mobile/apps/photos/lib/services/magic_cache_service.dart

@@ -2,7 +2,6 @@ import "dart:async";
 import "dart:convert";
 import "dart:io";
 
-import "package:computer/computer.dart";
 import "package:flutter/foundation.dart";
 import "package:flutter/widgets.dart";
 import "package:logging/logging.dart";
@@ -21,12 +20,12 @@ import "package:photos/models/search/hierarchical/magic_filter.dart";
 import "package:photos/models/search/search_types.dart";
 import "package:photos/service_locator.dart";
 import "package:photos/services/machine_learning/semantic_search/semantic_search_service.dart";
-import "package:photos/services/remote_assets_service.dart";
 import "package:photos/services/search_service.dart";
 import "package:photos/ui/viewer/search/result/magic_result_screen.dart";
 import "package:photos/utils/cache_util.dart";
 import "package:photos/utils/file_util.dart";
 import "package:photos/utils/navigation_util.dart";
+import "package:photos/utils/text_embeddings_util.dart";
 import "package:shared_preferences/shared_preferences.dart";
 
 class MagicCache {
@@ -119,8 +118,13 @@ GenericSearchResult? toGenericSearchResult(
   }
   if (!prompt.recentFirst) {
     enteFilesInMagicCache.sort((a, b) {
-      return fileIdToPositionMap[a.uploadedFileID!]!
-          .compareTo(fileIdToPositionMap[b.uploadedFileID!]!);
+      final aID = a.uploadedFileID;
+      final bID = b.uploadedFileID;
+      if (aID == null || bID == null) return 0;
+      final aPos = fileIdToPositionMap[aID];
+      final bPos = fileIdToPositionMap[bID];
+      if (aPos == null || bPos == null) return 0;
+      return aPos.compareTo(bPos);
     });
   }
   final String title = getLocalizedTitle(context, prompt.title);
@@ -168,7 +172,6 @@ GenericSearchResult? toGenericSearchResult(
 
 class MagicCacheService {
   static const _lastMagicCacheUpdateTime = "last_magic_cache_update_time";
-  static const _kMagicPromptsDataUrl = "https://discover.ente.io/v2.json";
 
   /// Delay is for cache update to be done not during app init, during which a
   /// lot of other things are happening.
@@ -178,7 +181,6 @@ class MagicCacheService {
   late final Logger _logger = Logger((MagicCacheService).toString());
 
   Future<List<MagicCache>>? _magicCacheFuture;
-  Future<List<Prompt>>? _promptFuture;
   final Set<String> _pendingUpdateReason = {};
   bool _isUpdateInProgress = false;
 
@@ -213,11 +215,7 @@ class MagicCacheService {
     if (!enableDiscover) {
       return;
     }
-    final updatedJSONFile = await RemoteAssetsService.instance
-        .getAssetIfUpdated(_kMagicPromptsDataUrl);
-    if (updatedJSONFile != null) {
-      queueUpdate("Prompts data updated");
-    } else if (lastMagicCacheUpdateTime <
+    if (lastMagicCacheUpdateTime <
         DateTime.now()
             .subtract(const Duration(hours: 12))
             .millisecondsSinceEpoch) {
@@ -247,10 +245,7 @@ class MagicCacheService {
       _isUpdateInProgress = true;
       final EnteWatch? w = kDebugMode ? EnteWatch("magicCacheWatch") : null;
       w?.start();
-      final magicPromptsData = await getPrompts();
-      w?.log("loadedPrompts");
-      final List<MagicCache> magicCaches =
-          await _nonEmptyMagicResults(magicPromptsData);
+      final List<MagicCache> magicCaches = await _nonEmptyMagicResults();
       w?.log("resultComputed");
       _magicCacheFuture = Future.value(magicCaches);
       await writeToJsonFile<List<MagicCache>>(
@@ -262,7 +257,6 @@ class MagicCacheService {
       await _resetLastMagicCacheUpdateTime();
       w?.logAndReset('done');
       _pendingUpdateReason.clear();
-      Bus.instance.fire(MagicCacheUpdatedEvent());
     } catch (e, s) {
       _logger.info("Error updating magic cache", e, s);
     } finally {
@@ -271,14 +265,6 @@ class MagicCacheService {
     }
   }
 
-  Future<List<Prompt>> getPrompts() async {
-    if (_promptFuture != null) {
-      return _promptFuture!;
-    }
-    _promptFuture = _readPromptFromDiskOrNetwork();
-    return _promptFuture!;
-  }
-
   Future<List<MagicCache>> getMagicCache() async {
     if (_magicCacheFuture != null) {
       return _magicCacheFuture!;
@@ -287,17 +273,6 @@ class MagicCacheService {
     return _magicCacheFuture!;
   }
 
-  Future<List<Prompt>> _readPromptFromDiskOrNetwork() async {
-    final String path =
-        await RemoteAssetsService.instance.getAssetPath(_kMagicPromptsDataUrl);
-    return Computer.shared().compute(
-      _loadMagicPrompts,
-      param: <String, dynamic>{
-        "path": path,
-      },
-    );
-  }
-
   Future<List<MagicCache>> _readResultFromDisk() async {
     _logger.info("Reading magic cache result from disk");
     final cache = await decodeJsonFile<List<MagicCache>>(
@@ -312,6 +287,7 @@ class MagicCacheService {
     if (file.existsSync()) {
       await file.delete();
     }
+    _magicCacheFuture = null;
   }
 
   Future<List<GenericSearchResult>> getMagicGenericSearchResult(
@@ -322,7 +298,30 @@ class MagicCacheService {
           kDebugMode ? EnteWatch("magicGenericSearchResult") : null;
       w?.start();
       final magicCaches = await getMagicCache();
-      final List<Prompt> prompts = await getPrompts();
+
+      // Load discover embeddings to get prompts
+      final discoverEmbeddings = await loadDiscoverEmbeddings();
+      if (discoverEmbeddings == null) {
+        _logger.severe("No discover embeddings available in assets");
+        throw Exception("No discover embeddings available in assets");
+      }
+
+      final prompts = <Prompt>[];
+      for (final entry in discoverEmbeddings.queryToPromptData.entries) {
+        final query = entry.key;
+        final promptData = entry.value;
+        prompts.add(
+          Prompt(
+            query: query,
+            title: promptData.title,
+            minScore: promptData.minScore,
+            minSize: promptData.minSize,
+            showVideo: promptData.showVideo ?? true,
+            recentFirst: promptData.recentFirst ?? false,
+          ),
+        );
+      }
+
       if (magicCaches.isEmpty) {
         w?.log("No magic cache found");
         return [];
@@ -375,35 +374,53 @@ class MagicCacheService {
     }
   }
 
-  static Future<List<Prompt>> _loadMagicPrompts(
-    Map<String, dynamic> args,
-  ) async {
-    final String path = args["path"] as String;
-    final File file = File(path);
-    final String contents = await file.readAsString();
-    final Map<String, dynamic> promptsJson = jsonDecode(contents);
-    final List<dynamic> promptData = promptsJson['prompts'];
-    return promptData
-        .map<Prompt>((jsonItem) => Prompt.fromJson(jsonItem))
-        .toList();
-  }
-
-  ///Returns non-empty magic results from magicPromptsData
-  ///Length is number of prompts, can be less if there are not enough non-empty
-  ///results
-  Future<List<MagicCache>> _nonEmptyMagicResults(
-    List<Prompt> magicPromptsData,
-  ) async {
+  Future<List<MagicCache>> _nonEmptyMagicResults() async {
     final TimeLogger t = TimeLogger();
     final results = <MagicCache>[];
     final List<int> matchCount = [];
-    final Map<String, double> queryToScore = {};
-    for (Prompt prompt in magicPromptsData) {
-      queryToScore[prompt.query] = prompt.minScore;
+
+    // Uncomment this code to generate embeddings and save them to a JSON file
+    // await generateAndSaveDiscoverEmbeddings();
+    // (from package:photos/utils/text_embeddings_util.dart)
+
+    // Load pre-computed discover embeddings from assets
+    final discoverEmbeddings = await loadDiscoverEmbeddings();
+    if (discoverEmbeddings == null) {
+      _logger.severe('Failed to load discover embeddings');
+      throw Exception('Failed to load discover embeddings');
     }
-    final clipResults =
-        await SemanticSearchService.instance.getMatchingFileIDs(queryToScore);
-    for (Prompt prompt in magicPromptsData) {
+
+    // Build the embeddings and score maps
+    final Map<String, List<double>> queryToEmbedding = {};
+    final Map<String, double> queryToScore = {};
+    final List<Prompt> prompts = [];
+
+    for (final entry in discoverEmbeddings.queryToPromptData.entries) {
+      final query = entry.key;
+      final promptData = entry.value;
+      final vector = discoverEmbeddings.queryToVector[query];
+
+      if (vector != null) {
+        queryToEmbedding[query] = vector.toList();
+        queryToScore[query] = promptData.minScore;
+        prompts.add(
+          Prompt(
+            query: query,
+            title: promptData.title,
+            minScore: promptData.minScore,
+            minSize: promptData.minSize,
+            showVideo: promptData.showVideo ?? true,
+            recentFirst: promptData.recentFirst ?? false,
+          ),
+        );
+      }
+    }
+
+    _logger.info('Using pre-computed discover embeddings from assets');
+    final clipResults = await SemanticSearchService.instance
+        .getMatchingFileIDsWithEmbeddings(queryToEmbedding, queryToScore);
+
+    for (final prompt in prompts) {
       final List<int> fileUploadedIDs = clipResults[prompt.query] ?? [];
       if (fileUploadedIDs.isNotEmpty) {
         results.add(

mobile/apps/photos/lib/services/smart_memories_service.dart

@@ -104,7 +104,7 @@ class SmartMemoriesService {
       );
 
       // Load pre-computed text embeddings from assets
-      final textEmbeddings = await loadTextEmbeddingsFromAssets();
+      final textEmbeddings = await loadMemoriesEmbeddings();
       if (textEmbeddings == null) {
         _logger.severe('Failed to load pre-computed text embeddings');
         throw Exception(

mobile/apps/photos/lib/utils/text_embeddings_util.dart

@@ -9,13 +9,14 @@ import "package:path_provider/path_provider.dart"
     show getExternalStorageDirectory;
 import 'package:photos/models/memories/clip_memory.dart';
 import 'package:photos/models/memories/people_memory.dart';
+import 'package:photos/models/ml/discover/prompt.dart';
 import "package:photos/services/machine_learning/ml_computer.dart"
     show MLComputer;
 
 final _logger = Logger('TextEmbeddingsUtil');
 
-/// Loads pre-computed text embeddings from assets
-Future<TextEmbeddings?> loadTextEmbeddingsFromAssets() async {
+/// Loads pre-computed memories text embeddings from assets
+Future<MemoriesEmbeddings?> loadMemoriesEmbeddings() async {
   try {
     _logger.info('Loading text embeddings from assets');
     final jsonString =
@@ -81,7 +82,7 @@ Future<TextEmbeddings?> loadTextEmbeddingsFromAssets() async {
     }
 
     _logger.info('Text embeddings loaded successfully from JSON assets');
-    return TextEmbeddings(
+    return MemoriesEmbeddings(
       clipPositiveVector: clipPositiveVector,
       peopleActivityVectors: peopleActivityVectors,
       clipMemoryTypeVectors: clipMemoryTypeVectors,
@@ -92,18 +93,94 @@ Future<TextEmbeddings?> loadTextEmbeddingsFromAssets() async {
   }
 }
 
-class TextEmbeddings {
+/// Container for all memories text embeddings
+class MemoriesEmbeddings {
   final Vector clipPositiveVector;
   final Map<PeopleActivity, Vector> peopleActivityVectors;
   final Map<ClipMemoryType, Vector> clipMemoryTypeVectors;
 
-  const TextEmbeddings({
+  const MemoriesEmbeddings({
     required this.clipPositiveVector,
     required this.peopleActivityVectors,
     required this.clipMemoryTypeVectors,
   });
 }
 
+/// Loads pre-computed discover text embeddings from assets
+Future<DiscoverEmbeddings?> loadDiscoverEmbeddings() async {
+  try {
+    _logger.info('Loading discover embeddings from assets');
+    final jsonString =
+        await rootBundle.loadString('assets/ml/discover_embeddings.json');
+    final data = json.decode(jsonString) as Map<String, dynamic>;
+
+    final promptsList = data['prompts'] as List;
+    final Map<String, Vector> queryToVector = {};
+    final Map<String, PromptData> queryToPromptData = {};
+
+    // Parse all discover embeddings
+    for (final promptJson in promptsList) {
+      final query = promptJson['query'] as String;
+      final vector = (promptJson['vector'] as List).cast<double>();
+      if (vector.isNotEmpty) {
+        queryToVector[query] = Vector.fromList(vector);
+        queryToPromptData[query] = PromptData(
+          title: promptJson['title'] as String,
+          minScore: (promptJson['minScore'] as num).toDouble(),
+          minSize: (promptJson['minSize'] as num).toDouble(),
+          showVideo: promptJson['showVideo'] as bool?,
+          recentFirst: promptJson['recentFirst'] as bool?,
+        );
+      }
+    }
+
+    // Check if we have all embeddings
+    if (queryToVector.isEmpty) {
+      _logger.warning('No discover embeddings found');
+      return null;
+    }
+
+    _logger.info(
+      'Discover embeddings loaded successfully with ${queryToVector.length} queries',
+    );
+    return DiscoverEmbeddings(
+      queryToVector: queryToVector,
+      queryToPromptData: queryToPromptData,
+    );
+  } catch (e, stackTrace) {
+    _logger.severe('Failed to load discover embeddings', e, stackTrace);
+    return null;
+  }
+}
+
+/// Container for discover/magic embeddings
+class DiscoverEmbeddings {
+  final Map<String, Vector> queryToVector;
+  final Map<String, PromptData> queryToPromptData;
+
+  const DiscoverEmbeddings({
+    required this.queryToVector,
+    required this.queryToPromptData,
+  });
+}
+
+/// Container for prompt metadata
+class PromptData {
+  final String title;
+  final double minScore;
+  final double minSize;
+  final bool? showVideo;
+  final bool? recentFirst;
+
+  const PromptData({
+    required this.title,
+    required this.minScore,
+    required this.minSize,
+    this.showVideo,
+    this.recentFirst,
+  });
+}
+
 /// Helper function to generate text embeddings and save them to a JSON file
 /// Run this once to generate the embeddings, then copy the output
 /// to assets/ml/text_embeddings.json
@@ -170,3 +247,161 @@ Future<void> generateAndSaveTextEmbeddings() async {
     '_generateAndSaveTextEmbeddings: Text embeddings generation complete! Copy the JSON output above to assets/ml/text_embeddings.json',
   );
 }
+
+/// Helper function to generate discover text embeddings and save them to a JSON file
+/// Run this once to generate the embeddings, then copy the output
+/// to assets/ml/discover_embeddings.json
+Future<void> generateAndSaveDiscoverEmbeddings() async {
+  // Hardcoded prompts from v2.json
+  final magicPromptsData = [
+    Prompt(
+      query: "identity documents and cards",
+      title: "Identity",
+      minScore: 0.20,
+      minSize: 0.0,
+      showVideo: false,
+    ),
+    Prompt(
+      query: "screenshots",
+      title: "Screenshots",
+      minScore: 0.18,
+      minSize: 0.0,
+      recentFirst: true,
+      showVideo: false,
+    ),
+    Prompt(
+      query: "receipts, bills and invoices",
+      title: "Receipts",
+      minScore: 0.20,
+      minSize: 0.0,
+      recentFirst: true,
+    ),
+    Prompt(
+      query: "screenshots of digital notes, photos of notebook and whiteboards",
+      title: "Notes",
+      minScore: 0.20,
+      minSize: 0.0,
+      recentFirst: true,
+    ),
+    Prompt(
+      query: "memes",
+      title: "Memes",
+      minScore: 0.18,
+      minSize: 0.0,
+      recentFirst: true,
+    ),
+    Prompt(
+      query: "business card",
+      title: "Visiting Cards",
+      minScore: 0.20,
+      minSize: 0.0,
+      recentFirst: true,
+      showVideo: false,
+    ),
+    Prompt(
+      query: "photo of toddler",
+      title: "Babies",
+      minScore: 0.20,
+      minSize: 0.0,
+      recentFirst: true,
+    ),
+    Prompt(
+      query: "photos of pets",
+      title: "Pets",
+      minScore: 0.20,
+      minSize: 0.0,
+      recentFirst: true,
+    ),
+    Prompt(
+      query: "selfie",
+      title: "Selfies",
+      minScore: 0.18,
+      minSize: 0.0,
+      recentFirst: true,
+    ),
+    Prompt(
+      query: "high resolution hd wallpaper",
+      title: "Wallpapers",
+      minScore: 0.18,
+      minSize: 0.0,
+      showVideo: false,
+    ),
+    Prompt(
+      query: "photo of food",
+      title: "Food",
+      minScore: 0.20,
+      minSize: 0.0,
+      recentFirst: true,
+    ),
+    Prompt(
+      query: "celebration with friends",
+      title: "Celebrations",
+      minScore: 0.20,
+      minSize: 0.0,
+      recentFirst: true,
+    ),
+    Prompt(
+      query: "photo of setting sun",
+      title: "Sunset",
+      minScore: 0.20,
+      minSize: 0.0,
+    ),
+    Prompt(
+      query: "photo of mountains or hills",
+      title: "Hills",
+      minScore: 0.20,
+      minSize: 0.0,
+    ),
+    Prompt(
+      query: "outdoor photo of trees, leaves, flowers",
+      title: "Greenery",
+      minScore: 0.20,
+      minSize: 0.0,
+    ),
+  ];
+
+  final Map<String, dynamic> embeddingsData = {
+    'version': '1.0.0',
+    'prompts': [],
+  };
+
+  final prompts = <Map<String, dynamic>>[];
+
+  for (final prompt in magicPromptsData) {
+    _logger.info('Generating embedding for discover prompt: ${prompt.title}');
+    final vector = await MLComputer.instance.runClipText(prompt.query);
+
+    final promptData = {
+      'query': prompt.query,
+      'title': prompt.title,
+      'minScore': prompt.minScore,
+      'minSize': prompt.minSize,
+      'vector': vector,
+    };
+
+    // Add optional fields
+    promptData['showVideo'] = prompt.showVideo;
+    promptData['recentFirst'] = prompt.recentFirst;
+
+    prompts.add(promptData);
+  }
+
+  embeddingsData['prompts'] = prompts;
+
+  // Convert to JSON and log it
+  final jsonString = const JsonEncoder.withIndent('  ').convert(embeddingsData);
+  dev.log(
+    '_generateAndSaveDiscoverEmbeddings: Generated discover embeddings JSON',
+  );
+
+  final tempDir = await getExternalStorageDirectory();
+  final file = File('${tempDir!.path}/discover_embeddings.json');
+  await file.writeAsString(jsonString);
+  dev.log(
+    '_generateAndSaveDiscoverEmbeddings: Saved discover embeddings to ${file.path}',
+  );
+
+  dev.log(
+    '_generateAndSaveDiscoverEmbeddings: Discover embeddings generation complete! Copy the JSON output above to assets/ml/discover_embeddings.json',
+  );
+}